A new approach to personal identification via hierarchical palmprint coding

This paper presents a new approach to personal identification using palmprints. To tackle the key issues such as palmprint feature extraction, representation, indexing, matching criteria and fast search for large data archives, we introduce, a hierarchical palmprint coding scheme to facilitate coarse-to-fine matching for efficient and effective palmprint classification. The proposed scheme consists of four levels: 1) a global geometry based feature for palmprint location (Level-1 feature), 2) a global texture based feature for palmprint grouping (Level-2 feature), 3) a local fuzzy "interest line" detection for palmprint selection (Level-3 feature), 4) a local texture energy vector for fine palmprint matching (Level-4 feature). In addition, we adopt a mask based statistical measurement to extract texture feature in palmprints, which is characterized by its high convergence of inner-palm similarities and good dispersion of inter-palm discrimination. Furthermore, we apply fuzzy set technology to detect dominant feature, points ("interest" points) and lines ("interest" lines) to guide matching. A wide range of palmprint images have been tested to demonstrate the feasibility and effectiveness of the proposed method.